17:43 <flamingradian[m]> Is there a chance that socinfo can get a stable userspace API?

17:51 <flamingradian[m]> apparently it's needed for sensors on xiaomi devices

18:14 <z3ntu> flamingradian You mean as in not being documented in Documentation/ABI/testing/sysfs-devices-soc? So "stable" instead of "testing" directory?

18:18 <flamingradian[m]> I mean the hardware_platform_subtype, platform_version, etc. that has been in debugfs

19:24 <aka_[m]> lumag: any idea how QOS SSPP feature work?

19:25 <aka_[m]> in case of 8976 i see exposed QOS_CTRL register but if QOS feature is set then it writes lut tables then enable ctrl

19:25 <aka_[m]> however there is no lut registers there

19:25 <aka_[m]> reading them from running mdp5 ends with zeros

19:30 <aka_[m]> guess im doing features of ();

19:34 <konradybcio> flamingradian: what exactly do you need?

19:47 <flamingradian[m]> The SLPI on xiaomi-beryllium requests the hardware_platform, hardware_platform_subtype, and platform_version socinfo values, and someone wrote an MR on HexagonRPCD to provide it (https://gitlab.com/sdm670-mainline/hexagonrpc/-/merge_requests/1/diffs#2d26bc67c8cc46998e8cf35a289236517c6e8d6c_165_171)

19:50 <flamingradian[m]> since it's via debugfs, I would like the locations to be stable so it doesn't break due to a change in the kernel

19:55 <konradybcio> flamingradian: do these values actually differ?

19:55 <konradybcio> as in, can you hardcode them for beryllium?

19:59 <flamingradian[m]> I can't tell, presumably not (but hardware_platform can definitely be hardcoded because it will always be "OEM")

20:00 <flamingradian[m]> *presumably no difference between same boards

20:02 <lumag> flamingradian[m]. Those are debugfs. Please don't depend on them.

20:02 <lumag> It should be safe to disable debugfs completely

20:04 <lumag> flamingradian[m], I think /sys/firmware/devicetree/base/compatible is the best call

20:15 <lumag> aka_[m], there is no per-pipe LUT on 8x76. See drivers/video/msm/mdss/mdss_mdp.c

20:16 <flamingradian[m]> lumag: ok, thanks, we will find an alternative

20:17 <lumag> It's there only on 8996/8937/8917/8953. Check for the MDSS_QOS_PER_PIPE_LUT

20:18 <lumag> flamingradian[m], wait... Do you need those values to identify the device or to set those values at runtime?

20:20 <lumag> If the former, then please use /sys/firmware/devicetree. At least it's closer to the ABI state.

20:20 <flamingradian[m]> lumag: we need to pass the values to the DSP fw

20:26 <lumag> flamingradian[m], you can hardcode hardware_platform and hardware_platform_subtype, those values won't ever change on an existing device.

20:27 <flamingradian[m]> ok, thanks for the info

20:28 <lumag> flamingradian[m], hardcode for the platform.

20:29 <lumag> other platforms might in theory have different values

20:29 <lumag> I'd need to check with pix-3...